1. Field of the Invention
This invention relates to plasma etching of platinum. More specifically, this invention provides an improved method for plasma etching of platinum for producing semiconductor integrated circuits containing platinum electrodes.
2. Description of the Prior Art
The implementation of digital information storage and retrieval is a common application of modern digital electronics. Memory size and access time serve as a measure of progress in computer technology. Quite often storage capacitors are employed as memory array elements. As the state of the art has advanced, small-feature-size high density dynamic random access memory (DRAM) devices require storage capacitors of larger capacitance having high dielectric constant materials. The high dielectric constant materials or ferroelectric materials are made primarily of sintered metal oxide and contain a substantial amount of very reactive oxygen. In the formation of capacitors with such ferroelectric materials or films, the electrodes must be composed of materials with least reactivity to prevent oxidation of the electrodes which would decrease the capacitance of storage capacitors. Therefore, precious metals, such as platinum (Pt), palladium (Pd), etc., are preferred metals used in the manufacture of capacitors for high density DRAM.
Among the possible precious metals for capacitor electrodes, platinum has emerged as an attractive candidate because it is inert to oxidation and is known to have a leakage current (&lt;10.sup.-9 amps/cm.sup.2) lower than other electrodes such as RuO.sub.2 and Pd. Platinum also is a good conductor.
In the prior art, platinum etching has been conducted by means of isotropic etching, such as wet etching with aqua regia, or by anisotropic etching, such as ion milling with Ar gas or by other means. Because of the nature of isotropic etching, using wet etching with aqua regia causes deteriorated processing accuracy. The grade of precision in isotropic etching is not high enough for fine pattern processing. Therefore, it is difficult to perform submicron patterning of platinum electrodes due to its isotropic property. Furthermore, a problem with ion milling (i.e. anisotropic etching) occurs because the etching speed on platinum, which is to form the electrode, is too slow for mass production.
In order to increase processing accuracy in etching platinum, research and development has been quite active, particularly in the area of etching platinum by means of a dry etching process where etchant gases (e.g., Cl.sub.2, HBr, O.sub.2, etc.) are used. The following prior art is representative of the state of art with respect to etching platinum with a plasma of etching gases.
U.S. Pat. No. 5,492,855 to Matsumoto et al. discloses a semiconductor device manufacturing method, wherein an insulation layer, a bottom electrode Pt layer, a dielectric film and a top electrode Pt layer are provided on top of a substrate having already-completed circuit elements and wiring, and then, a capacitor is formed by selectively dry etching the bottom electrode Pt layer after selectively dry etching the top electrode Pt layer and the dielectric film. The manufacturing method uses a gas containing an S component as etching gas for Pt etching, or an etching gas containing S component as an additive gas; and also it implants S into the Pt layer before the Pt dry etching process by means of ion implantation to compose a S and Pt compound, and then dry etches the Pt compound thus composed.
U.S. Pat. No. 5,527,729 to Matsumoto et al. discloses process steps to form on a substrate in which circuit elements and wirings, etc., are already shaped, an insulation layer, a first metal layer, a dielectric film and a second metal layer. A top electrode and a capacitance film are formed by dry etching the second metal layer and the dielectric film. A bottom electrode is formed by dry etching the first metal layer. The etching gas for dry etching the second metal layer is a mixed gas containing hydrogen halide (e.g. HBr) and oxygen, having a ratio of oxygen against the total of hydrogen halide and oxygen set at about 10%-35%. The etching gas is also taught as a gas containing hydrocarbon, such as chloroform. Matsumoto et al. employs a silicon oxide layer as the insulation layer on the substrate, and a platinum layer or palladium layer as the first and second metal layers. Dry etching of the second metal layer and dielectric film is conducted in a low pressure region not higher than about 5 Pa, where the etching speed is high. Matsumoto et al. further teaches that where a mixed gas of hydrogen halide and oxygen is used as the etching gas, the etching speed on the silicon oxide layer can be made sufficiently low relative to that on the second metal layer made of a platinum layer or a palladium layer; in this way, the excessive etching of the silicon oxide layer underlying the first metal layer is avoided, and damage to the circuit elements and wiring, etc. underneath the silicon oxide layer can be prevented. Furthermore according to Matsumoto et al, the ratio of etching speed of the platinum and dielectric material to the resist can be increased by lowering the etching speed on the resist. Therefore, etching of the platinum and dielectric material may be conducted by using a mask of normal lay-thickness resist (generally speaking, about 1.2 .mu.m to about 2.0 .mu.m thick), instead of using a conventional thick-layer resist (about 3 .mu.m and thicker).
Chou et al. in an article entitled "Platinum Metal Etching in a Microwave Oxygen Plasma", J. Appl. Phys. 68 (5), Sep. 1, 1990, pages 2415-2423, discloses a study to understand the etching of metals in both plasma and chemical systems. The study found that the etching of platinum foils in an oxygen plasma generated in a flow-type microwave system and that very rapid etching (.about.6.ANG./s) took place even at low power inputs (200 W). The principal plasma parameters, including oxygen atom concentration, ion concentration, and electron temperature, were measured by Chou et al. as a function of distance below the microwave coupler. These were correlated to the rate of foil etching, which decreased with increasing distance from the coupler. On the basis of these correlations Chou et al. formulated a simple mechanistic model. The study by Chou et al. further found that the etching of platinum in an oxygen plasma jet results from the concomitant action of oxygen atoms and high energy electrons.
Nishikawa et al. in an article entitled "Platinum Etching and Plasma Characteristics in RF Magnetron and Electron Cyclotron Resonance Plasmas", Jpn. J. Appl. Phys., Vol. 34 (1995), pages 767-770, discloses a study wherein the properties of platinum etching were investigated using both rf magnetron and electron cyclotron resonance (ECR) plasmas, together with measurement of the plasma parameters (neutral concentration, plasma density, etc.). Nishikawa et al. performed experiments in Cl.sub.2 plasmas over a pressure ranging from 0.4 to 50 mTorr. In rf magnetron plasmas, the etch rate of Pt was constant at the substrate temperature of from 20 to 160.degree. C. The etch rate and the plasma electron density increased with gas pressure decreasing from 50 to 5 mTorr. In ECR plasmas for rf power of 300 W, Nishikawa et al. found that the etch rate of Pt was almost constant (.about.100 nm/min) with gas pressure decreasing from 5 to 0.4 mTorr, while the plasma electron density gradually increased with decreasing gas pressure. The study by Nishikawa et al. discusses these experimental results with respect to the relationship between the etch yield and the ratio of neutral Cl.sub.2 flux and ion flux incident on the substrate.
Yokoyama et al. in an article entitled "High-Temperature Etching of PZT/Pt/TiN Structure by High-Density ECR Plasma", Jpn. J. Appl. Phys., Vol. 34 (1995), pages 767-770, discloses a study wherein micron patterning technologies for the PZT/Pt/TiN/Ti structure with a spin on glass (SOG) mask are demonstrated using a high-density electron cyclotron resonance (ECR) plasma and a high substrate temperature above 300.degree. C. A 30%-Cl.sub.2 /Ar gas was used to etch a lead zirconate titanate (PZT) film. No deposits remained, which resulted in an etched profile of more than 80.degree.. A 40%-O.sub.2 /Cl.sub.2 gas was used to etch a Pt film. The etching was completely stopped at the Ti layer. 30-nm-thick deposits remained on the sidewall. They were removed by Yokoyama et al. after dipping in hydrochloric acid. The etched profile of a Pt film was more than 80.degree.. The Ti/TiN/Ti layer was etched with pure Cl.sub.2 gas. The size shift from the SOG mask was less than 0.1 .mu.m. Yokoyama et al. did not detect any interdiffusion between SOG and PZT by transmission electron microscopy and energy dispersive x-ray spectroscopy (TEM-EDX) analysis.
Yoo et al. in an article entitled "Control of Etch Slope During Etching of Pt in Ar/Cl.sub.2 /O.sub.2 Plasmas", Jpn. J. Appl. Phys., Vol. 35 (1996), pages 2501-2504, teaches etching of Pt patterns of the 0.25 .mu.m design rule at 20.degree. C. using a magnetically enhanced reactive ion etcher (MERIE). Yoo et al. found that a major problem of etching with a MERIE was the redeposition of the etch products onto the pattern sidewall, making it difficult to reduce the pattern size. In both cases separately using a photoresist mask and an oxide mask, the redeposits of the etch products onto the sidewall were reduced by the addition of Cl.sub.2 to Ar, although the etched slope was lowered to 45.degree.. The redeposits were removed by an HCl cleaning process.
Kotecki in an article entitled "High-K Dielectric Materials for DRAM Capacitors", Semiconductor International, November 1996, pages 109-116, the potential advantages of incorporating high-dielectric materials into a storage capacitor of a dynamic random access memory (DRAM) are described and the requirements of the high dielectric layer are reviewed as they relate to use in a simple stack capacitor structure suitable for the gigabit generation. Kotecki teaches that when considering the use of high-dielectric materials in a stack capacitor structure, the following issues need to be addressed: electrode patterning, high-dielectric material/barrier interaction, electrode/high-dielectric material interaction, surface roughness (e.g. hilocking, etc.), step coverage, high-dielectric material uniformity (e.g. thickness, composition, grain size/orientation, etc.), and barrier (e.g. O.sub.2 and Si diffusion, conductivity, contact resistance and interactions, etc.). Various materials and combinations of materials were studied by Kotecki for use with perovskite dielectrics including the noble metals (i.e. Pt, Ir, Pd) and conductive metal oxides (i.e. IrO.sub.2 and RuO.sub.2). The work function of these materials, their ability to be patterned by dry etching, the stability of the surface with regards to surface roughening and their suitability in a semiconductor fabricator are listed by Kotecki in the following Table I:
TABLE I Comparison of the Properties of Various Electrode Materials Suitable for Use with Perovskite Dielectrics Material Work Dry Surface Deposition Selection Function Etch Stability Method Pt 5.6-5.7 difficult potential problem sputtering Ru 4.7 easy/dangerous potential problem sputtering RuO.sub.2 /Ru easy/dangerous good reactive sputtering Ir 5.0-5.8 difficult good sputtering IrO.sub.2 /Ir difficult good reactive sputtering Pd 5.1-5.6 difficult ? sputtering
Kotecki further teaches in the article entitled "High-K Dielectric Materials for DRAM Capacitors" that one of the major problems which needs to be overcome with respect to the manufacturing of DRAM chips using capacitors is the problem of electrode patterning. There are minimal volatile species produced during the dry etching of the noble metal electrodes such as Pt, Ru, Pd and Ir. Since the etch mechanism is primarily by physical sputtering, even during a RIE process, fences are typically formed on the sides of the photoresist. To eliminate the problem of fencing, it is possible to etch the fence layer and erode the sides of the photoresist during the etch process which leads to "clean" metal structures but with sloping sidewall angles and a loss of control over critical feature sizes. As the dimension of the feature shrinks to 0.18 .mu.m or below, only limited tapering of the sidewall angle can be tolerated. Kotecki presents in the following Table II some of the high-dielectric materials which have been considered for use in a DRAM capacitor, the various methods which can be used to form the films, and the range of reported permittivites:
TABLE II A Comparison of Various High-Dielectric Materials and Method for Formation and Dielectric Constants Material Deposition Methods .epsilon..sub.T (thin films) SrTiO.sub.3 MOCVD, ECR-CVD, sol-gel, 90-240 sputtering, PLD (Ba, Sr)TiO.sub.3 MOCVD, ECR-CVD, sol-gel, 160-600 sputtering, PLD PLT MOCVD, sol-gel, sputtering, PLD 400-900 PZT and PLZT MOCVD, sol-gel, sputtering, PLD &gt;1000
Milkove et al. reported in a paper entitled "New Insight into the Reactive Ion Etching of Fence-Free Patterned Platinum Structures" at the 43rd Symposium of AVS, October 1996, Philadelphia, Pa., that an investigation was undertaken to characterize the time progression of the Pt etch process during the reactive ion etching (RIE) of fence-free patterned structures. The experiment by Milkove et al. consisted of coprocessing two oxidized Si wafers possessing identical 2500 .ANG. thick Pt film layers, but different photoresist (PR) mask thicknesses. Etching was suspended at 20, 40, 60 and 80% of the full etch process in order to cleave off small pieces of wafer for analysis by a scanning electron microscopy (SEM). Using Cl.sub.2 -based RIE conditions known to produce fence-free etching for 2500 .ANG. thick film layers, Milkove et al. discovered that a severe fence actually coats the PR mask during the first 20% of the etch process. As the etch continues the fence structure evolves, achieving a maximum height and width followed by progressive recession until disappearing completely prior to process endpoint. The data from Milkove et al. shows that the final profile of an etched Pt structure possess a functional dependence on the initial thickness and slope of the PR mask, as well as on the initial thickness of the Pt layer. Milkove et al. further reported in the paper entitled "New Insight Into The Reactive Ion Etching of Fence-free Patterned Platinum Structures" that the observed behavior of the transient fence provides the strongest evidence to date supporting the existence of a chemically assisted physical sputtering component associated with the RIE of Pt films in halogen-based plasmas.
Keil et al. teaches in an article entitled "The Etching of Platinum Electrodes for PZT Based Ferroelectric Devices", Electrochemical Society Proceedings, Vol. 96-12 (1996), pages 515-520, that the technical difficulties of fabricating capacitors employing platinum Pt etching is most often dominated by sputtering processes. While oxygen and/or various gaseous chlorides or fluorides are used to chemically enhance the etch process, the products of both etch mechanisms are usually of low volatility and tend to redeposit on the wafer. After etching, large wall-like structures extend up from the edges of the Pt region. These wall-like structures are frequently referred to as "veils" or "fences" or "rabbit ears" and can reach lengths which are more than double the thickness of the Pt film to which they are attached. The existence of such structures makes useful deposition of the PZT layer impossible. Keil et al. further teaches that even when one is able to attenuate redeposition to the point where only small "nub" like features are present, the high electric fields which will form at such "nubs" enhances the likelihood for dielectric breakdown. Although process conditions can be found which result in either low redeposition or even no redeposition, they most often also give an unacceptably tapered platinum profile angle. Keil et al. observed that redeposition becomes more severe as process conditions are pushed toward those which give increasingly vertical sidewalls. While a post etch wet clean in a solvent bath is frequently used, the heavy redeposition which attends the pursuit of vertical sidewalls regularly renders this approach minimally effective.
The foregoing prior art illustrates that generally a clean vertical dense area profile and CD (critical dimension) control of the etch profiles are critical factors for successful plasma etching of 1-Gbit (and beyond) DRAM ferroelectric devices possessing platinum electrodes. Redeposition and profile control are found to be strongly interlinked. Optimization of both profile angle and redeposition requires a tradeoff between the two. Where as vigorous post etch cleaning (e.g. wet cleaning with acid, mechanical polishing, etc.) can relieve some of the need to achieve a deposition free plasma etch, such post etch cleaning does not possess the accuracy that is desired as the platinum electrode itself is typically eroded and/or deteriorated by currently known post etch cleaning methods.
Therefore, what is needed and what has been invented is a method for etching a platinum electrode layer to produce a high density integrated circuit semiconductor device having platinum electrodes with a high degree (i.e., .gtoreq.85.degree.) of platinum profile anisotropy. What is further needed and what has been invented is a semiconductor device including a plurality of platinum electrodes having a platinum profile equal to or greater than about 85.degree. and separated by a distance equal to or less than about 0.3 .mu.m with each electrode having a critical dimension (e.g., a width) equal to or less than about 0.3 .mu.m.